3
$\begingroup$

I am reading a paper which refers to a maximally skewed stable distribution $F(x;1,-1,\pi/2,0)$ . Is there an efficient way to sample from this distribution?

If $X$ has distribution $F(x;\alpha,\beta,\gamma,\delta)$, then when $\alpha=1$ its characteristic function $\phi(\theta) = \mathbb{E}(e^{-i\theta X})$, $\theta \in \mathbb{R}$ is given by

$$\phi(\theta) = e^{\gamma(-|\theta| - i\theta\beta (2/\pi)\log{|\theta|})+i\delta\theta}.$$

$\endgroup$

2 Answers 2

1
$\begingroup$

there are the packages stabledist and FMStable that will do this for you.

$\endgroup$
2
  • $\begingroup$ Thank you. I am not sure this package exactly covers the $\alpha=1$ case does it? It seems you would end up taking the limit as $\alpha \to 1$ which is a numerically unstable approach. $\endgroup$
    – Simd
    May 20, 2014 at 12:09
  • $\begingroup$ FMStable allows you to set $1-\alpha$ explicitly and claims high accuracy. $\endgroup$ May 20, 2014 at 12:20
1
$\begingroup$

Table 1 of http://arxiv.org/abs/0908.3961v2 contains an explicit algorithm for this.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.